Background
Post-traumatic stress disorder (PTSD) is a well recognised psychiatric disorder that can occur following a major traumatic event. Characteristic symptoms include re-experiencing phenomena such as nightmares and recurrent distressing thoughts of the event, avoidance and numbing of general responsiveness such as trying not to talk about or be reminded of the traumatic event, experiencing detachment and estrangement from other people and hyperarousal symptoms including sleep disturbance, increased irritability and hypervigilance. PTSD is a relatively common condition. The National Co-morbidity Survey (Kessler 1995) found that 7.8% of 5,877 American adults had suffered from PTSD at some time in their lives. When data were examined from individuals who had been exposed to a traumatic event rates of PTSD varied according to type of stressor. For example, physical assaults amongst women led to a lifetime prevalence of 29% and combat experience amongst men to a lifetime prevalence of 39%. It is apparent that PTSD causes much suffering and that developing effective interventions is important.
Psychological interventions have been advocated as being effective in the treatment of PTSD since its conception. Various forms of psychological treatment have been used including exposure therapy, cognitive therapy, stress inoculation training, psychodynamic psychotherapy and eye movement desensitisation and reprocessing (EMDR) (Foa 2000). Exposure therapy usually involves asking the subject to relive the trauma imaginally. This is often done by creating a detailed present tense account of exactly what happened, making an audio tape recording of it and asking the individual to listen to this over and over again. Another form of exposure therapy involves exposing subjects to cues associated with the traumatic event (for example graded re-exposure to car travel following a road traffic accident). Trauma-focused cognitive therapy involves helping the individual to identify distorted thinking patterns regarding themselves, the traumatic incident and the world. Individuals are encouraged to challenge their thoughts by weighing up available evidence and through the utilization of various techniques by the therapist including specific questioning that leads the individual to challenge distorted views. EMDR involves the PTSD sufferer focusing on a traumatic image, thought, emotion and a bodily sensation whilst receiving bilateral stimulation most commonly in the form of eye movements. Psychodynamic psychotherapy focuses on integrating the traumatic experience into the life experience of the individual as a whole. Often childhood issues are felt to be important.
The psychological treatments described and a variety of others have their advocates, but much of this advocacy is based on anecdotal evidence only. All the treatments have a theoretical basis as to why they might work, but their true effectiveness in reducing symptoms or their potential adverse consequences is not really known. Solomon 1992 reviewed the treatment literature and concluded that most of the available studies had some methodological shortcomings and that there was a need for further evaluation. A more recently published meta-analysis included more randomised controlled trials (Sherman 1998) and practice guidelines from the International Society for Traumatic Stress Studies (Foa 2000) added to these. However, this topic has not yet been subjected to a systematic review adhering to the Cochrane Collaboration guidelines.
Objectives
To perform a systematic review of randomised controlled trials of psychological treatments for PTSD following the guidelines of the Cochrane Collaboration. The efficacy of psychological treatments in comparison with control conditions and other psychological treatments will be determined using clinician rated symptoms of PTSD as the main outcome measure.
Methods
Criteria for considering studies for this review
Types of studies
Any randomised controlled trial that considered one or more defined psychological treatments to reduce traumatic stress symptoms in comparison with a placebo, other control (e.g. usual care or waiting list control) or alternative psychological treatment condition was included. All studies must have been completed and analysed by October 2004 for inclusion. Sample size, language and publication status was not used to determine whether or not a study should be included.
Types of participants
Any individual suffering from traumatic stress symptoms with a duration of symptoms of three months or more. At least 70% of participants had to be diagnosed as suffering from PTSD according to DSM or ICD criteria This review considered studies of adults only. There was no restriction on the basis of severity of PTSD symptoms, type of traumatic event or comorbidity (including major depressive disorder), however, PTSD had to be considered the primary diagnosis for individuals to be included.
Types of interventions
This review considered any psychological treatment designed to reduce symptoms of PTSD. The review has now been updated to include eye movement desensitisation and reprocessing (EMDR). Other Cochrane Collaboration reviews have considered brief psychological interventions for treating immediate trauma-related symptoms and preventing PTSD (Rose 2004) and pharmacological treatments for the treatment of PTSD (Stein 2004).
The following eligible treatment categories were identified.
a. Trauma focused cognitive behavioural therapy (TFCBT) - Any psychological treatment delivered individually that predominantly used trauma focused cognitive, behavioural or cognitive-behavioural techniques. This category included exposure therapy.
b. Stress management/relaxation - Any psychological treatment delivered individually that predominantly used non-trauma focused cognitive, behavioural or cognitive-behavioural techniques.
c. TFCBT Group Therapy - Any approach delivered in a group setting that predominantly used trauma focused cognitive, behavioural or cognitive-behavioural techniques.
d. Non-trauma focused CBT group therapy - Any approach delivered in a group that predominantly used non-trauma focused cognitive, behavioural or cognitive-behavioural techniques.
e. Other psychological treatment - Any psychological treatment delivered individually that predominantly used non-trauma focused techniques that would not be considered cognitive, behavioural or cognitive-behavioural techniques. This category included non-directive counselling, psychodynamic therapy and hypnotherapy.
f. Eye movement desensitisation and reprocessing (EMDR).
g. Wait list/usual care - These ranged from no intervention at all to undefined psychological input and/or drug treatment that was not fully described.
Types of outcome measures
Categorical and continuous variables were used:
Primary outcome measure:
1. The primary outcome measure was severity of clinician rated traumatic stress symptoms using a standardised measure such as the Clinician Administered PTSD Symptom Scale (Blake 1995).
Secondary outcome measures:
1. Severity of self-reported traumatic stress symptoms using a standardised measure such as the Impact of Event Scale (Horowitz 1979).
2. Severity of depressive symptoms using scales such as the Beck Depression Inventory (Beck 1961).
3. Severity of anxiety symptoms using scales such as the Spielberger State Trait Anxiety Inventory (Spielberger 1973).
4. Dropout rates.
5. PTSD diagnosis after treatment.
6. Any adverse effects, e.g. increased PTSD symptoms.
Search methods for identification of studies
This involved a systematic review of a variety of sources using methods described by the Cochrane Collaboration in August 2005. Computerised databases were searched using the Cochrane optimal RCT search strategy combined with the following key words: PTSD, trauma, cognitive, behavioural, exposure, EMDR, psychological, psychotherapy, psychodynamic, stress inoculation, relaxation, anxiety management.
Databases - Medline, clinpsych, psychlit, Embase, Pilots (a specialized PTSD database maintained by the National PTSD Centre in the USA), Trials Register of the Cochrane Depression, Anxiety and Neurosis Group, lilacs, psynebs, sociofile.
Hand Searches - Journal of Traumatic Stress, ISTSS Treatment Guidelines (Foa 2000)
Reference Lists - Of studies identified in the search
Internet Search - Of known websites and discussion fora
Personal Communication - The main source of personal communication was with the NICE guidelines development group who kindly shared the results of their searches and communications with the following people: Arnoud Arntz & Merel Kindt, Richard Bryant, Willi Butollo, Claude Chemtob, Judith Cohen, Mark Creamer, Jonathan Davidson, Enrique Echeburua, Paul Emmelkamp, Edna Foa, Chris Freeman, Berthold Gersons, Louise Humprheys, Terry Keane, Dean Kilpatrick, Edward Kubany, Brett Litz, Andreas Maercker, Charles Marmar, Sandy McFarlane, Thomas Mellman, Lars-Goran Öst, Michael Otto, Roger Pitman, Mark Pollack, Patti Resick, David Riggs, Sue Rose, Barbara Rothbaum, Joe Ruzek, Patricia White, Paula Schnurr, Matt Friedman, Arieh Shalev, Dan Stein, Nick Tarrier, Agnes van der Minnen, Simon Wessely and Rachel Yehuda.
Abstracts/Dissertations - from meetings of the European and International Societies of Traumatic Stress Studies.
Data collection and analysis
Applying selection criteria - Abstracts of all potential trials identified through the search strategy were independently read by the two reviewers. If an abstract was felt to possibly represent a RCT the full report was fully read by each reviewer independently to determine if the trial met the inclusion criteria.
Extracting data - Spreadsheets were designed to capture data which was then entered using the Review Management software. Information extracted included demographic details of participants, details of the traumatic event, the randomisation process, the interventions used and outcome data.
Assessment of methodological quality - This combined the standard approach described in the Cochrane Handbook which considers randomisation, allocation concealment and intention to treat with a quality score from a predetermined scale (Moncrieff 2001). This scale considers 23 different methodological criteria and assigns scores to them on a 0-2 scale giving a maximum possible total of 46. The criteria included in the scale are objectives and specification of main outcomes a priori, sample size, follow up duration, power calculation, method of allocation, allocation concealment, clear description of treatment and adjunctive treatment, blinding of subjects, representative sample recruitment, use of diagnostic criteria, exclusion criteria and number of exclusions and refusals, description of sample demographics, blinding of assessor, assessment of compliance with treatments, details of side-effects, record of number and reasons for withdrawal by group, outcome measures described clearly or use of validated instruments, information on comparability and adjustment for differences in analysis, inclusion of withdrawals in analysis, presentation of results with inclusion of data for reanalysis of main outcomes, appropriate statistical analysis, conclusions justified and declaration of interests.
The Cochrane criteria and other scale were scored by both reviewers independently. Disagreements were discussed between the reviewers in order to make a final decision regarding the quality score of the study.
Analyses
The following information about the identified trials was presented:
1. Included RCTs and their year of publication.
2. Excluded studies with reason for exclusion.
3. The characteristics of participants.
4. The nature of the psychological treatment and control condition considered.
5. The methodological quality of the RCTs using the methods described above.
6. The pooled effects of the overall effects in individual trials.
The following tables were presented:
1. A table of characteristics of the RCTs included in the review.
2. A table summarising the methodological quality of the RCTs included in the review.
Calculation of treatment effects:
The data were summarised and pooled effects calculated using RevMan 4.1software. Continuous outcomes were analysed as standardised mean differences (SMDs) to allow for ease of comparison across studies. It was decided to use relative risk as the main categorical outcome measure as this is more widely understood than odds ratios in medical practice.
Comparisons:
The following comparisons were used (i) psychological treatment vs waitlist or usual care control; (ii) psychological treatment vs another psychological treatment.
Choice of Method for Pooling Data
Data were pooled from more than one study using a fixed effects meta-analysis, except where heterogeneity was present in which case a random-effects model was used as described below.
Heterogeneity
To check for heterogeneity between studies, both the I squared test of heterogeneity and the chi-squared test of heterogeneity (p < .10), as well as visual inspection of the forest plots were used. An I squared of less than 30% was taken to indicate mild heterogeneity and a fixed effects model was used to synthesise the results. An I squared of more than 50% was taken as notable heterogeneity. In this case, an attempt was made to explain the variation. If studies with heterogeneous results were found to be comparable, a random effects model was used to summarise the results. An I squared of 30% to 50% was taken to indicate moderate heterogeneity. In this case, both the chi-squared test of heterogeneity and a visual inspection of the forest plot were used to decide between a fixed and random effects model.
Clinical heterogeneity subgroup analyses were performed for studies that only included females and studies that did not include Vietnam veterans for the primary outcome comparison of TFCBT vs wait list/usual care. All trials that scored above 25 on the Moncrieff 2001 scale were considered "higher quality studies". Studies that scored below 26 on the Moncrieff 2001 scale were considered "lower quality studies". Sensitivity analyses were performed for higher quality studies and lower quality studies.
Results
Description of studies
See: Characteristics of included studies; Characteristics of excluded studies.
Trials excluded
See excluded trials table.
Studies were excluded if they did not satisfy the inclusion criteria. Other reasons for excluding specific studies were less than three months following trauma and therefore PTSD had not been present for three months or more (Echeburua 1996; Frank 1988), treatment for anger only (Chemtob 1997), relaxation treatments only with no comparison (Walsh) and comparison of two CBT techniques only (Tarrier 1999; Paunovic 2001).
Trials included
See included trials table
Thirty-three different trials fulfilled the inclusion criteria.
Patient selection
See characteristics of trials included.
The study populations were varied and not directly comparable (i.e. there was significant clinical heterogeneity). Six studies included male Vietnam veterans only (Carlson 1998, Cooper 1989, Jensen 1994, Keane 1989, Peniston 1991 and Schnurr 2003), twelve studies considered female assault (mainly sexual assault) survivors (Classen 2001, Cloitre 2002, Echeburua 1997, Foa 1991, Foa 1999, Krakow 2001, Kubany 2003, Kubany 2004, Resick 2002, Rothbaum 1997, Rothbaum 2005 and Zlotnick 1997), two studies included only road traffic accident survivors (Blanchard 2003 and Fecteau 1999), one study was of refugees (Neuner 2004), one of police officers (Gersons 2000) and eleven studies included individuals from various traumas including road traffic accidents, assaults, bereavement and industrial accidents (Brom 1989, Bryant 2003, Devilly 1999, Ehlers 2003, Ironson 2002, Lee 2002, Marcus 1997/2004 (single study with follow-up), Marks 1998, Power 2002, Scheck 1998 and Vaughan 1994). The majority of participants satisfied the criteria for a DSM diagnosis of PTSD although some studies included individuals with traumatic stress symptoms who did not fulfil the full DSM criteria. The Vietnam veteran studies were largely from samples of individuals already in care. Other studies often advertised for their participants or used referrals to an established traumatic stress service.
Cultural Setting:
United States of America (23 studies), Australia (2 studies), United Kingdom (3 studies), The Netherlands (2 studies), Germany (1 study) and Canada (2 studies).
Sample size:
The number of patients randomised to the trials ranged from 16 (Cooper 1989 and Peniston 1991) to 360 (Schnurr 2003). Four studies included sample sizes of over 100 (Schnurr 2003 (360), Resick 2002 (121), Krakow 2001 (114) and Brom 1989 (112)).
Time post Trauma:
All studies included individuals at least three months following the trauma. The range was large , from 3 months to over 30 years. There was often a wide range of times since trauma included in individual studies.
Interventions:
In order to present the results in a meaningful way it was decided to pool data that used a similar theoretical methodology. This resulted in the establishment of seven groups - TFCBT, stress management, trauma focused group CBT, non-trauma focused group CBT, psychodynamic therapy, hypnotherapy and supportive counselling. Because of the existence of only one trial in each of the last three groups it was decided to pool these as "other therapies" for the purposes of this review.
Trauma focused cognitive behavioural therapy - Twenty-two studies considered TFCBT - Blanchard 2003, Brom 1989, Bryant 2003, Cloitre 2002, Cooper 1989, Echeburua 1997, Ehlers 2003, Fecteau 1999, Foa 1991, Foa 1999, Gersons 2000, Keane 1989, Kubany 2003, Kubany 2004, Marks 1998, Neuner 2004, Peniston 1991, Power 2002, Resick 2002, Rothbaum 2005, Taylor 2003 and Vaughan 1994.
Stress management - Seven studies considered stress management - Carlson 1998, Echeburua 1997, Foa 1991, Foa 1999, Marks 1998, Taylor 2003 and Vaughan 1994.
Group trauma focused CBT - Four studies considered group trauma focused CBT - Classen 2001, Krakow 2001, Schnurr 2003 and Zlotnick 1997.
Other therapies - Four studies considered other therapies - Blanchard 2003, Brom 1989, Bryant 2003 and Foa 1991.
Eye Movement Desensitisation and Reprocessing - Twelve studies considered eye movement desensitisation and reprocessing - Carlson 1998, Devilly 1999, Ironson 2002, Jensen 1994, Lee 2002, Marcus 1997/2004, Power 2002, Rothbaum 1997, Rothbaum 2005, Scheck 1998, Taylor 2003 and Vaughan 1994.
Comparisons:
The included trials compared (i) psychological treatment vs waitlist or usual care control (some studies allowed the control group to receive pharmacological treatments and/or psychological treatments that were not being considered specifically); (ii) psychological treatment vs other psychological treatment.
The following specific comparisons were made:
a. TFCBT versus waitlist/usual care - Blanchard 2003, Brom 1989, Cloitre 2002, Cooper 1989, Ehlers 2003, Fecteau 1999, Foa 1991, Foa 1999, Gersons 2000, Keane 1989, Kubany 2003, Kubany 2004, Marks 1998, Peniston 1991, Power 2002, Resick 2002, Rothbaum 2005 and Vaughan 1994.
b. Stress management versus waitlist/usual care - Carlson 1998, Foa 1991, Foa 1999 and Vaughan 1994.
c. Other therapies versus waitlist/usual care - Blanchard 2003, Brom 1989 and Foa 1991.
d. Group CBT versus waitlist/usual care - Classen 2001, Krakow 2001 and Zlotnick 1997.
e. TFCBT versus stress management - Echeburua 1997, Foa 1991, Foa 1999, Marks 1998, Taylor 2003 and Vaughan 1994.
f. TFCBT versus other therapies - Blanchard 2003, Brom 1989, Bryant 2003, Foa 1991 and Neuner 2004.
g. Stress management versus other therapy - Foa 1991.
h. Group TFCBT versus group non trauma focused CBT - Schnurr 2003.
i. EMDR versus waitlist/usual care - Carlson 1998, Jensen 1994, Power 2002, Rothbaum 1997, Rothbaum 2005 and Vaughan 1994.
j. EMDR versus TFCBT - Devilly 1999, Ironson 2002, Lee 2002, Power 2002, Rothbaum 2005, Taylor 2003 and Vaughan 1994.
k. EMDR versus stress management - Carlson 1998, Taylor 2003 and Vaughan 1994.
l. EMDR versus other therapy - Marcus 1997/2004 and Scheck 1998.
Risk of bias in included studies
Randomisation
Most studies did not provide full details of the method of allocation and some bias was believed to be possible from the description in 26 studies. In seven studies the method of allocation was felt to be appropriate with no bias possible (Bryant 2003, Schnurr 2003, Krakow 2001, Marks 1998, Resick 2002, Scheck 1998 and Vaughan 1994).
Allocation concealment
Most studies did not provide full details of the method of randomisation and therefore concealment was unclear in 27 studies (Blanchard 2003, Brom 1989, Bryant 2003, Carlson 1998, Classen 2001, Cloitre 2002, Devilly 1999, Echeburua 1997, Ehlers 2003, Foa 1999, Gersons 2000, Ironson 2002, Jensen 1994, Keane 1989, Kubany 2004, Lee 2002, Marcus 1997/2004, Marks 1998, Peniston 1991, Resick 2002, Rothbaum 1997, Rothbaum 2005, Scheck 1998, Schnurr 2003, Taylor 2003, Vaughan 1994, Zlotnick 1997). There was evidence of adequate concealment in the Power 2002 study. In six studies randomisation concealment was inadequate, for example using the roll of a die or a list of randomised numbers (Cooper 1989, Fecteau 1999, Foa 1991, Krakow 2001, Kubany 2003, Neuner 2004).
Blinding
In common with all studies of psychological treatment a double blind methodology is virtually impossible as it is clear to the subject what treatment they are receiving. However, a well designed study should have ensured blinding of the assessor of outcome measures. This was performed in 20 studies (Blanchard 2003, Bryant 2003, Cloitre 2002, Ehlers 2003, Fecteau 1999, Foa 1999, Gersons 2000, Krakow 2001, Kubany 2003, Kubany 2004, Marks 1998, Neuner 2004, Peniston 1991, Power 2002, Resick 2002, Rothbaum 2005, Scheck 1998, Schnurr 2003, Taylor 2003, Vaughan 1994) but not present in the other studies. In no studies was the blinding complimented by a test for the integrity of it.
Loss to follow-up
This was fully reported with reasons by group in eleven studies (Blanchard 2003, Ehlers 2003, Fecteau 1999, Gersons 2000, Ironson 2002, Krakow 2001, Neuner 2004, Peniston 1991, Rothbaum 1997, Taylor 2003, Vaughan 1994). In three studies this was not recorded (Jensen 1994, Keane 1989, Marcus 1997/2004). In the other studies withdrawals were recorded without reasons by group.
Moncrieff et al (2001) assessment:
The scores for each item and total scores for all the studies are shown in the methodological quality table. It is important to view the items separately as it is likely that some studies with higher scores had significant methodological shortcomings. The overall quality of the studies was variable. Several studies had significant flaws as is illustrated by the table. One trend was that the earlier studies tended to have lower quality scores than the more recent ones. Sixteen studies had a score of 26 or more including 13 of the 15 studies published in the 21st century. Sixteen studies scored below 25 including all four studies published in the 1980s. Only three studies (Krakow 2001, Power 2002, Schnurr 2003) scored over 30.
There was rarely any measure of treatment fidelity and only one study (Taylor 2003) provided details of any side effects of treatment although this was only information regarding worsening of specific symptoms on the main outcome measure. In several studies the conclusions were only partially justified from the results obtained. A strength of the majority of the studies was having clear objectives but sample sizes were small and the follow-up period was limited. Fifteen studies had follow-up periods of six months or more (Bryant 2003, Carlson 1998, Classen 2001, Echeburua 1997, Foa 1999, Krakow 2001, Kubany 2004, Marcus 2004, Marks 1998, Neuner 2004, Peniston 1991, Power 2002, Resick 2002, Rothbaum 2005 and Schnurr 2003). Power calculations were rarely reported and it is apparent that many of the studies were underpowered. The treatments delivered were reasonably well described although there was limited testing of treatment fidelity. The majority of studies used well validated outcome measures although there was considerable variation in the actual measures used.
TFCBT
The TFCBT study scores ranged from 17 (Brom 1989) to 32 (Power 2002). The overall quality was variable and has been further explored in a sensitivity analysis reported in the results section.
There were several specific aspects of individual studies that need to be considered when interpreting the results. Blanchard 2003 included individuals with "severe sub-syndromal PTSD" defined as individuals who did not fully meet either the re-experiencing, avoidance or hyperarousal criteria but did meet all other DSMIV criteria. In the Brom 1989 study, 83 (74%) had experienced bereavement as the trauma and the period between therapeutic sessions was unclear. As part of the assessment interview "confrontation therapy" was used apparently to determine the reaction to the traumatic event. Clearly this may have affected outcome. In the Cooper 1989 study most individuals finishing the imaginal flooding continued to receive both standard individual and group therapies. Both usual care group and treatment group subjects received a standard component treatment (individual and group) designed for PTSD. This comprised weekly sessions of one hour evaluating symptoms and background of PTSD with an educational component. Two hour weekly group sessions focussed on a number of problem areas including PTSD symptoms using group problem-solving, current life problems and group support. Clearly the usual care group received significant treatment in this study.
Foa 1991 excluded assaults by a spouse or family member. Foa 1999 was one of the best studies methodologically although how subjects were recruited was unclear and the number of drop-outs was not specified. Gersons 2000 too was a well designed study but restricted to male police officers. The types of trauma were not specified although all fulfilled Criterion A of the DSMIV classification of PTSD. Keane 1989 did not describe the severity or type of trauma, nor the time between trauma and study. Treatment and waiting list groups continued to receive medication throughout the trial. An unknown number of the usual care/waiting list control group subjects continued to attend programmes for veterans or to see a psychiatrist and there was no data on the degree of involvement or treatment given in this group during the study. This is likely to have reduced the validity of this study and specifically the ability to detect a difference in effectiveness between the two groups. In common with several of the studies of Vietnam veterans this study appears to have been of men with chronic, probably treatment resistant PTSD symptoms with a relatively poor prognosis and compared an active treatment against a usual care control group who were also receiving significant ongoing treatment.
The Marks 1998 study was quite strong methodologically. Exclusions included those who had had past treatment with cognitive therapy, suggesting a bias in favour of those whose symptomatology or illness may have been less severe and clearly in contrast with the methodology employed in several of the Vietnam veteran studies leading to a likely better outcome. Unfortunately there was a high attrition rate and the later follow-up data were often on very small groups. The Peniston 1991 study suffers from low sample size and chronic Vietnam veteran PTSD sufferers - ten of the sixteen were inpatients. There were no drop-outs and no detail on missed sessions. The Resick 2002 study was very strong methodologically with a large sample size. Ehlers 2003 study was strong methodologically but suffered from a small sample size.
Vaughan's study included individuals from a range of traumas of whom 22% did not satisfy the DSM III R criteria for PTSD. No homework was given in the eye movement desensitisation group whereas the applied muscle relaxation and image habituation therapy groups were required to complete homework
Stress management:
The quality scores of these studies varied from 21 (Vaughan 1994) to 29 (Foa 1999). Issues concerning the Foa 1991, Foa 1999, Marks 1998 and Vaughan 1994 studies have been discussed above. The Carlson 1998 study suffered from a small sample size. In the Echeburua 1997 study outcome assessments were performed by the therapists themselves and the treatment was not manualised.
Other therapies
These studies all included TFCBT as an intervention as well as other therapies. The quality scores varied from 17 (Brom 1989) to 29 (Bryant 2003).
Group TFCBT
The studies of group TFCBT included the two of the studies with the highest quality scores (Krakow 2001 31, Schnurr 2003 37) and two of the studies that scored least (Zlotnick 1997 18, Classen 2001 20). The Krakow 2001 treatment focused on nightmares and did not specifically deal with other phenomena of PTSD which may have impacted on the results. In the Zlotnick 1997 study the duration of symptoms was not apparent. All subjects were receiving individual psychotherapy in addition to the group intervention and medication was being prescribed throughout . There were seven (29%) drop-outs in the treatment group with no reasons given for dropping out. However those not completing had higher scores on the pre-treatment PTSD symptom scale and the Dissociative Experience Scale. The presence of PTSD at the end of the study was estimated from the Davidson trauma scale questionnaire as opposed to a structured interview post treatment. Schnurr 2003 had the largest sample size of all the studies but unfortunately there was no wait-list or other non-active treatment control group which makes interpretation very difficult.
EMDR
The quality scores of EMDR studies ranged from 18 (Jensen 1994) to 32 (Power 2002). Unfortunately most studies included only small sample sizes, the maximum number of individuals who had received EMDR in any analysis was 109. Most of the studies assessed fidelity of the EMDR but were less diligant regarding the fidelity of other treatments delivered. There was also great variability between studies with regards to the number of EMDR sessions (from Ironson 2002 - 1-3 to Devilly 1999 - 12).
Effects of interventions
The full results are contained in the tables and are summarised below.
1. TFCBT/Exposure therapy versus waitlist/usual care
Clinician rated PTSD symptoms:
Fourteen studies considered this outcome with a total of 649 individuals. There was significant statistical heterogeneity between these trials (Chi square = 88.89; p<0.00001: I square = 85.4%) and a random effects model was used to pool the data. The TFCBT group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.40 (-1.89 to -0.91)).
Self reported PTSD symptoms:
Nine studies considered this outcome with a total of 428 individuals. There was significant statistical heterogeneity between these trials (Chi square = 29.7; p = 0.0002: I square = 73.1%) and a random effects model was used to pool the data. The TFCBT group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.68 (-2.14 to -1.22)).
Depression:
Fourteen studies considered this outcome with a total of 625 individuals. There was significant statistical heterogeneity between these trials (Chi square = 69.16; p<0.00001: I square = 81.2%) and a random effects model was used to pool the data. The TFCBT group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.26 (-1.69 to -0.82)).
Anxiety:
Eleven studies considered this outcome with a total of 415 individuals. There was no significant statistical heterogeneity between these trials. The TFCBT group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -0.99 (-1.2 to -0.78)).
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
Fifteen studies with a total of 861 individuals recorded whether individuals left the study early for any reason by group. There was no significant statistical heterogeneity between these trials. The TFCBT group did significantly worse than the waitlist/usual care group (RR (95% CI) = 1.42 (1.05, 1.94)).
PTSD diagnosis after treatment:
Fifteen studies with a total of 756 individuals reported this outcome. There was significant statistical heterogeneity between these trials (Chi square = 62.88; p<0.00001: I square = 77.7%) and a random effects model was used to pool the data. The TFCBT group did significantly better than the waitlist/usual care group (RR (95% CI) = 0.44 (0.34, 0.57)).
2. Stress management versus waitlist/usual care:
Clinician rated PTSD symptoms:
Four studies considered this outcome with a total of 86 individuals. There was no significant statistical heterogeneity between these trials. The stress management group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.14 (-1.62 to -0.67)).
Self reported PTSD symptoms:
One study considered this outcome with a total of 24 individuals. There was no statistically significant difference between the stress management group and the waitlist/usual care group immediately after treatment (SMD (95% CI) = 0.33 (-0.47 to 1.14)).
Depression:
Four studies considered this outcome with a total of 109 individuals. There was no significant statistical heterogeneity between these trials. The stress management group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -0.73 (-1.12 to -0.33)).
Anxiety:
Three studies considered this outcome with a total of 82 individuals. There was no significant statistical heterogeneity between these trials. The stress management group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -0.77 (-1.23 to -0.31)).
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
Four studies with a total of 121 individuals recorded whether individuals left the study early for any reason by group. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the stress management group and the waitlist/usual care group (RR (95% CI) = 2.19 (0.71, 6.73)).
PTSD diagnosis after treatment:
Three studies with a total of 121 individuals reported this outcome. There was significant statistical heterogeneity between these trials (Chi square = 8.63; p = 0.03: I square = 65.2%) and a random effects model was used to pool the data. The stress management group did significantly better than the waitlist/usual care group (RR (95% CI) = 0.64 (0.47, 0.87)).
3. Other therapies versus waitlist/usual care:
Clinician rated PTSD symptoms:
Two studies considered this outcome with a total of 72 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the other therapies group and the waitlist/usual care group immediately after treatment (SMD (95% CI) = -0.43 (-0.9 to 0.04)).
Self reported PTSD symptoms:
Two studies considered this outcome with a total of 132 individuals. There was no significant statistical heterogeneity between these trials. The other therapies group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -0.61 (-0.98 to -0.24)).
Depression:
Two studies considered this outcome with a total of 72 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the other therapies group and the waitlist/usual care group immediately after treatment (SMD (95% CI) = -0.25 (-0.71 to 0.22)).
Anxiety:
Three studies considered this outcome with a total of 153 individuals. There was no significant statistical heterogeneity between these trials. The other therapies group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -0.48 (-0.82 to -0.14)).
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
Three studies with a total of 166 individuals recorded whether individuals left the study early for any reason by group. There was no significant statistical heterogeneity between these trials. The other therapies group did significantly worse than the waitlist/usual care group (RR (95% CI) = 3.82 (1.19, 12.29)).
PTSD diagnosis after treatment:
Three studies with a total of 166 individuals reported this outcome. There was significant statistical heterogeneity between these trials (Chi square = 8.72; p = 0.01: I square = 77.1%) and a random effects model was used to pool the data. There was no difference between the other therapies and the waitlist/usual care group (RR (95% CI) = 0.79 (0.53, 1.18)).
4. Group TFCBT versus waitlist/usual care:
Clinician rated PTSD symptoms:
One study considered this outcome with a total of 45 individuals. The group TFCBT group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -0.72 (-1.14 to -0.31)).
Self reported PTSD symptoms:
Two studies considered this outcome with a total of 71 individuals. There was no significant statistical heterogeneity between these trials. The group TFCBT group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -0.71 (-1.20 to -0.22)).
Depression:
No studies considered this outcome.
Anxiety:
No studies considered this outcome.
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
Three studies with a total of 271 individuals recorded whether individuals left the study early for any reason by group. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the group TFCBT group and the waitlist/usual care group (RR (95% CI) = 1.00 (0.64, 1.56)).
PTSD diagnosis after treatment:
One study with a total of 48 individuals reported this outcome. There was no significant difference between the group TFCBT group and the waitlist/usual care group (RR (95% CI) = 0.56 (0.31, 1.01)).
5. TFCBT/Exposure therapy versus stress management:
Clinician rated PTSD symptoms:
Six studies considered this outcome with a total of 239 individuals. There was significant statistical heterogeneity between these trials (Chi square = 11.25; p = 0.05: I square = 55.6%) and a random effects model was used to pool the data. There was no statistically significant difference between the TFCBT group and the stress management group immediately after treatment (SMD (95% CI) = -0.27 (-0.71 to 0.16)). At 2-5 month follow-up five studies considered this outcome with a total of 127 individuals. There was no significant statistical heterogeneity between these trials. The TFCBT group did significantly better than the stress management group at 2-5 month follow-up (SMD (95% CI) = -0.48 (-0.84 to -0.12)).
Self reported PTSD symptoms:
Three studies considered this outcome with a total of 127 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the TFCBT group and the stress management group immediately after treatment (SMD (95% CI) = -0.37 (-0.74 to 0.01)). At 2-5 month follow-up two studies considered this outcome with a total of 54 individuals. The TFCBT group did significantly better than the stress management group at 2-5 month follow-up (SMD (95% CI) = -0.44 (-0.99 to 0.10)).
Depression:
Five studies considered this outcome with a total of 161 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the TFCBT group and the stress management group immediately after treatment (SMD (95% CI) = -0.25 (-0.57 to 0.08)). At 2-5 month follow-up five studies considered this outcome with a total of 147 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the TFCBT group and the stress management group at 2-5 month follow-up (SMD (95% CI) = -0.28 (-0.62 to 0.06)).
Anxiety:
Four studies considered this outcome with a total of 127 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the TFCBT group and the stress management group immediately after treatment (SMD (95% CI) = -0.12 (-0.49 to 0.26)). At 2-5 month follow-up five studies considered this outcome with a total of 117 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the TFCBT group and the stress management group at 2-5 month follow-up (SMD (95% CI) = -0.19 (-0.58 to 0.20)).
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
Six studies with a total of 284 individuals recorded whether individuals left the study early for any reason by group. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the TFCBT group and the stress management group (RR (95% CI) = 1.17 (0.69, 2.0)).
PTSD diagnosis after treatment:
Six studies with a total of 284 individuals reported this outcome. There was no significant statistical heterogeneity between these trials. There was a statistically significant difference between the TFCBT group and the stress management group (RR (95% CI) = 0.78 (0.61, 0.99)).
6. TFCBT/Exposure therapy versus other therapies
Clinician rated PTSD symptoms:
Three studies considered this outcome with a total of 120 individuals. There was no significant statistical heterogeneity between these trials. The TFCBT group did significantly better than the other therapies group immediately after treatment (SMD (95% CI) = -0.81 (-1.19 to -0.42)). At 3 month follow-up two studies considered this outcome with a total of 70 individuals. There was no significant statistical heterogeneity between these trials. The TFCBT group did significantly better than the other therapies group at 3 month follow-up (SMD (95% CI) = -0.65 (-1.13 to -0.16)). One trial reported this outcome at 6 to 9 month follow-up and again found that the TFCBT group did significantly better than the other therapies group (SMD (95% CI) = -1.85 (-2.59 to -1.11)).
Self reported PTSD symptoms:
Three studies considered this outcome with a total of 176 individuals. There was significant statistical heterogeneity between these trials (Chi square = 21.90; p<0.0001: I square = 90.9%) and a random effects model was used to pool the data. The TFCBT group did significantly better than the other therapies group immediately after treatment (SMD (95% CI) = -1.18 (-2.32 to -0.03)). At 2-5 month follow-up two studies considered this outcome with a total of 131 individuals. There was significant statistical heterogeneity between these trials (Chi square = 4.43; p = 0.04: I square = 77.4%) and a random effects model was used to pool the data. There was no significant difference between the TFCBT and the other therapies group at 2-5 month follow-up (SMD (95% CI) = -0.28 (-1.04 to 0.48)). One trial reported this outcome at 6 to 9 month follow-up and again found that the TFCBT group did significantly better than the other therapies group (SMD (95% CI) = -1.72 (-2.45 to -1.00)).
Depression:
Three studies considered this outcome with a total of 120 individuals. There was no significant statistical heterogeneity between these trials. The TFCBT group did significantly better than the other therapies group immediately after treatment (SMD (95% CI) = -0.65 (-1.03 to -0.28)). At 2-5 month follow-up two studies considered this outcome with a total of 72 individuals. There was no significant statistical heterogeneity between these trials. The TFCBT group did significantly better than the other therapies group at 2-5 month follow-up (SMD (95% CI) = -0.53 (-1.00 to -0.05)).One trial reported this outcome at 6 to 9 month follow-up and again found that the TFCBT group did significantly better than the other therapies group (SMD (95% CI) = -1.08 (-1.74 to -0.42)).
Anxiety:
Four studies considered this outcome with a total of 197 individuals. There was significant statistical heterogeneity between these trials (Chi square = 12.85; p = 0.005: I square = 76.7%) and a random effects model was used to pool the data. There was no significant difference between the TFCBT and the other therapies group immediately after treatment (SMD (95% CI) = -0.47 (-1.11 to 0.17)). At 2-5 month follow-up three studies considered this outcome with a total of 149 individuals. There was no significant statistical heterogeneity between these trials. There was no significant difference between the TFCBT and the other therapies group at 2-5 month follow-up (SMD (95% CI) = -0.27 (-0.60 to 0.07)). One trial reported this outcome at 6 to 9 month follow-up and again found that the TFCBT group did significantly better than the other therapies group (SMD (95% CI) = -1.18 (-1.85 to -0.51)).
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
Five studies with a total of 290 individuals recorded whether individuals left the study early for any reason by group. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the TFCBT group and the other therapies group (RR (95% CI) = 1.14 (0.68, 1.9)).
PTSD diagnosis after treatment:
Five studies with a total of 286 individuals reported this outcome. There was no significant statistical heterogeneity between these trials. There was a statistically significant difference between the between the TFCBT and the other therapies group (RR (95% CI) = 0.71 (0.56, 0.89)).
7. Stress management versus other therapies
Clinician rated PTSD symptoms:
One study considered this outcome with a total of 25 individuals. The stress management group did significantly better than the other therapies group immediately after treatment (SMD (95% CI) = -1.22 (-2.09 to -0.35)). At 3 month follow-up one study considered this outcome with a total of 18 individuals. There was no significant difference between the stress management and the other therapies group at 3 month follow-up (SMD (95% CI) = -0.38 (-1.31 to 0.55)).
Self reported PTSD symptoms:
No studies considered this outcome.
Depression:
One study considered this outcome with a total of 25 individuals. There was no significant difference between the stress management and the other therapies group immediately after treatment (SMD (95% CI) = -0.51 (-1.31 to 0.30)). At 3 month follow-up one study considered this outcome with a total of 18 individuals. There was no significant difference between the stress management and the other therapies group at 3 month follow-up (SMD (95% CI) = -0.48 (-1.42 to 0.46)).
Anxiety:
One study considered this outcome with a total of 25 individuals. There was no significant difference between the stress management and the other therapies group immediately after treatment (SMD (95% CI) = -0.51 (-1.32 to 0.29)). At 3 month follow-up one study considered this outcome with a total of 18 individuals. There was no significant difference between the stress management and the other therapies group at 3 month follow-up (SMD (95% CI) = -0.68 (-1.64 to 0.28)).
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
One study with a total of 31 individuals recorded whether individuals left the study early for any reason by group. There was no statistically significant difference between the stress management group and the waitlist/usual care group (RR (95% CI) = 0.82 (0.20, 3.46)).
PTSD diagnosis after treatment:
One study with a total of 31 individuals reported this outcome. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the stress management group and the waitlist/usual care group (RR (95% CI) = 0.58 (0.30, 1.11)).
8. Group TFCBT versus group non-TF CBT
Clinician rated PTSD symptoms:
One study considered this outcome with a total of 325 individuals. There was no significant difference between the group TFCBT and non-trauma-focused CBT groups (SMD (95% CI) = -0.12 (-0.34 to 0.10)).
Self reported PTSD symptoms:
No studies considered this outcome.
Depression:
No studies considered this outcome.
Anxiety:
No studies considered this outcome.
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
One study with a total of 360 individuals recorded whether individuals left the study early for any reason by group. There was no statistically significant difference between the group TFCBT and non-trauma-focused CBT groups (RR (95% CI) = 1.38 (1.0, 1.9)).
PTSD diagnosis after treatment:
One study with a total of 360 individuals reported this outcome. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the group TFCBT and non-trauma-focused CBT groups (RR (95% CI) = 0.98 (0.83, 1.16)).
9. EMDR versus waitlist/usual care
Clinician rated PTSD symptoms:
Five studies considered this outcome with a total of 162 individuals. There was no significant statistical heterogeneity between these trials. The EMDR group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.51 (-1.87 to -1.15)).
Self reported PTSD symptoms:
Five studies considered this outcome with a total of 156 individuals. There was significant statistical heterogeneity between these trials (Chi square = 27.85; p < 0.0001: I square = 85.6%) and a random effects model was used to pool the data. There was no statistically significant difference between the EMDR and waitlist/usual care groups immediately after treatment (SMD (95% CI) = -1.07 (-2.04 to -.10)).
Depression:
Five studies considered this outcome with a total of 160 individuals. There was no significant statistical heterogeneity between these trials. The EMDR group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.48 (-1.84 to -1.12)).
Anxiety:
Five studies considered this outcome with a total of 156 individuals. There was no significant statistical heterogeneity between these trials. The EMDR group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.10 (-1.45 to -0.76)).
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
Six studies with a total of 217 individuals recorded whether individuals left the study early for any reason by group. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the EMDR and waitlist/usual care groups (OR (95% CI) = 1.33 (0.64 to 2.74)).
PTSD diagnosis after treatment:
Six studies with a total of 209 individuals reported this outcome. There was significant statistical heterogeneity between these trials (Chi square = 52.61; p<0.00001: I square = 90.5%) and a random effects model was used to pool the data. The EMDR group did significantly better than the waitlist/usual care group (RR (95% CI) = 0.47(0.25 to 0.85)).
10. EMDR versus TFCBT/ Exposure therapy:
Clinician rated PTSD symptoms:
Six studies considered this outcome with a total of 187 individuals. There was significant statistical heterogeneity between these trials (Chi square = 16.51; p = 0.006: I square = 69.7%) and a random effects model was used to pool the data. There was no statistically significant difference between the EMDR group and the TFCBT group immediately after treatment (SMD (95% CI) = 0.03 (-0.5 to 0.55)). At 2-5 month follow-up three studies considered this outcome with a total of 76 individuals. There was no statistically significant difference between the EMDR and TFCBT groups at 2-5 month follow-up (SMD (95% CI) = -0.14 (-0.60 to 0.32)).
Self reported PTSD symptoms:
Seven studies considered this outcome with a total of 206 individuals. There was significant statistical heterogeneity between these trials (Chi square = 13.33; p = 0.04: I square = 55%) and a random effects model was used to pool the data. There was no statistically significant difference between the EMDR group and the TFCBT group immediately after treatment (SMD (95% CI) = -0.17 (-0.59 to 0.26)). At 2-5 month follow-up five studies considered this outcome with a total of 111 individuals. There was no statistically significant difference between the EMDR and TFCBT groups at 2-5 month follow-up (SMD (95% CI) = -0.01 (-0.39 to 0.37)).
Depression:
Seven studies considered this outcome with a total of 206 individuals. There was significant statistical heterogeneity between these trials (Chi square = 23.99; p = 0.0005: I square = 75%) and a random effects model was used to pool the data. There was no statistically significant difference between the EMDR group and the TFCBT group immediately after treatment (SMD (95% CI) = -0.32 (-0.90 to 0.26)). At 2-5 month follow-up five studies considered this outcome with a total of 111 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the EMDR group and the TFCBT group at 2-5 month follow-up (SMD (95% CI) = -0.09 (-0.47 to 0.29)).
Anxiety:
Four studies considered this outcome with a total of 136 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the EMDR group and the TFCBT group immediately after treatment (SMD (95% CI) = -0.08 (-0.42 to 0.26)). At 2-5 month follow-up two studies considered this outcome with a total of 48 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the EMDR group and the TFCBT group at 2-5 month follow-up (SMD (95% CI) = 0.24 (-0.33 to 0.81)).
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
Seven studies with a total of 268 individuals recorded whether individuals left the study early for any reason by group. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the EMDR group and the TFCBT group (RR (95% CI) = 0.83(0.55 to 1.26 )).
PTSD diagnosis after treatment:
Six studies with a total of 260 individuals reported this outcome. There was significant statistical heterogeneity between these trials (Chi square = 14.38; p = 0.03: I square = 58.3%) and a random effects model was used to pool the data. There was no statistically significant difference between the EMDR group and the TFCBT group (RR (95% CI) = 1.11(0.68 to 1.81)).
11. EMDR versus stress management therapy:
Clinician rated PTSD symptoms:
Two studies considered this outcome with a total of 53 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the EMDR group and the stress management therapy group immediately after treatment (SMD (95% CI) = -0.35 (-0.90 to 0.19)). At 2-5 month follow-up three studies considered this outcome with a total of 71 individuals. The EMDR group did significantly better than the stress management therapy group immediately after treatment (SMD (95% CI) = -0.59 (-1.08 to -0.09)).
Self reported PTSD symptoms:
Three studies considered this outcome with a total of 75 individuals. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the EMDR group and the stress management therapy group immediately after treatment (SMD (95% CI) = -0.40 (-0.86 to 0.06)). At 2-5 month follow-up three studies considered this outcome with a total of 75 individuals. The EMDR group did significantly better than the stress management therapy group immediately after treatment (SMD (95% CI) = -0.52 (-0.98 to -0.05)).
Depression:
Three studies considered this outcome with a total of 75 individuals. There was no significant statistical heterogeneity between these trials. The EMDR group did significantly better than the stress management therapy group immediately after treatment (SMD (95% CI) = -0.67 (-1.14 to -0.20)). At 2-5 month follow-up three studies considered this outcome with a total of 75 individuals. There was no statistically significant difference between the EMDR group and the stress management therapy group at 2-5 month follow-up (SMD (95% CI) =
-0.23 (-0.70 to 0.23)).
Anxiety:
Two studies considered this outcome with a total of 45 individuals. There was no significant statistical heterogeneity between these trials. The EMDR group did significantly better than the stress management therapy group immediately after treatment (SMD (95% CI) = -0.75 (-1.36 to -0.13)). At 2-5 month follow-up two studies considered this outcome with a total of 45 individuals. There was no statistically significant difference between the EMDR group and the stress management therapy group at 2-5 month follow-up (SMD (95% CI) =
-0.42 (-2.21 to 1.37)).
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
Three studies with a total of 84 individuals recorded whether individuals left the study early for any reason by group. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the EMDR group and the stress mangement therapy group (RR (95% CI) = 1.03 (0.37 to 2.88)).
PTSD diagnosis after treatment:
Three studies with a total of 84 individuals reported this outcome. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the EMDR group and the stress management therapy group (RR (95% CI) = 0.69 (0.46 to 1.04)).
12. EMDR versus other therapies:
Clinician rated PTSD symptoms:
No studies formally considered this outcome.
Self reported PTSD symptoms:
Two studies considered this outcome with a total of 124 individuals. There was no significant statistical heterogeneity between these trials. The EMDR group did significantly better than the other therapies group immediately after treatment (SMD (95% CI) = -0.84 (-1.21 to -0.47)).
Depression:
Two studies considered this outcome with a total of 127 individuals. There was no significant statistical heterogeneity between these trials. The EMDR group did significantly better than the other therapies group immediately after treatment (SMD (95% CI) = -0.67 (-1.03 to -0.32)).
Anxiety:
Two studies considered this outcome with a total of 126 individuals. There was no significant statistical heterogeneity between these trials. The EMDR group did significantly better than the other therapies group immediately after treatment (SMD (95% CI) = -0.72 (-1.08 to -0.36)).
Adverse effects:
No studies formally considered adverse effects.
Dropouts:
Two studies with a total of 127 individuals recorded whether individuals left the study early for any reason by group. There was no significant statistical heterogeneity between these trials. There was no statistically significant difference between the EMDR group and the other therapies group (RR (95% CI) = 1.48 (0.26 to 8.54)).
PTSD diagnosis after treatment:
One study with a total of 67 individuals reported this outcome. The EMDR group did significantly better than the other therapies group (RR (95% CI) = 0.40 (0.19 to 0.84)).
Clinical heterogeneity subgroup analyses
In order to explore clinical heterogeneity, two subgroup analyses were performed for the primary outcome measure, i.e. clinician rated PTSD symptoms, for the TFCBT versus waitlist/usual care comparison. Thirteen studies had considered this outcome with a total of 609 individuals and the initial SMD was -1.36 (95% CI = -1.88 to -0.84), suggesting that the TFCBT group did significantly better than the waitlist/usual care group immediately after treatment. There was significant statistical heterogeneity between these trials (Chi square = 86.62; p<0.00001: I square = 86.1%).
Female only studies
Seven studies considered this outcome with a total of 404 females. The TFCBT group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.94 (-2.53 to -1.34)), demonstrating a larger difference in favour of TFCBT than in the overall analyses. Seven studies with mixed gender populations and a total of 145 inidividuals reported this outcome. Although the TFCBT group still did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -0.83 (-0.61 to -0.41)), the observed difference between groups was much reduced. The observed statistically significant heterogeneity remained following these subgroup analyses, although was much reduced in the mixed gender subgroup (Chi square = 14.52; p = 0.02: I square = 58.7%).
Studies not considering Vietnam veterans
Thirteen studies considered this outcome with a total of 625 individuals. The TFCBT group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.49 (-1.99 to -0.99)), demonstrating little difference from the overall analyses. Excluding this one trial made no difference to the observed statistically significant heterogeneity.
Sensitivity analyses
In order to explore the impact of methodological quality, a sensitivity analysis was performed for the primary outcome measure, i.e. clinician rated PTSD symptoms, for the TFCBT versus waitlist/usual care comparison.
Thirteen studies had considered this outcome with a total of 609 individuals and the initial SMD was -1.36 (95% CI = -1.88 to -0.84), suggesting that the TFCBT group did significantly better than the waitlist/usual care group immediately after treatment. The studies were divided into higher and lower quality studies. Nine higher quality studies considered this outcome with a total of 493 individuals. The TFCBT group again did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.61 (-2.16 to -1.06)). Five lower quality studies considered this outcome with a total of 156 individuals. Once more, the TFCBT group did significantly better than the waitlist/usual care group immediately after treatment (SMD (95% CI) = -1.02 (-1.84 to -0.20)), although in the lower quality studies the observed difference between groups was reduced. The observed statistically significant heterogeneity remained in each of these subgroup analyses.
Publication bias
All the studies identified for this review were published or were accepted for publication, and many of the trials were undertaken relatively recently. The potential effects of publication bias were explored using funnel plots. Two funnel plots were constructed using data from the TFCBT versus waitlist/usual care comparison, one involving continuous data on the primary outcome (clinician-rated PTSD symptoms - see Figure 1), and the second involving dichotomous data on a secondary outcome (PTSD diagnosis after treatment - see Figure 2). These funnel plots both show that the smaller studies may tend to report larger differences between TFCBT and Waitlist/Usual Care, and both suggest an absence of studies demonstrating no difference or a difference in favour of Waitlist/Usual care. It is therefore possible that, due to the greater likelihood of publication of positive studies, the true difference between groups is smaller than is suggested by this review.
Discussion
Trauma focused cognitive behavioural therapy
There was good evidence that TFCBT was better than wait list/usual care in reducing traumatic stress symptoms and additionally associated symptoms of depression and anxiety. It is possible that this may be stronger than suggested by the data, as in several studies the wait list/usual care group received some contact and the expectation that they would be treated which may have been therapeutic. However, there it is also possible that wait list groups do worse than usual care groups because they do not expect to improve until they receive the active intervention. The overall standardised mean difference for traumatic stress symptoms post treatment represents an effect size generally accepted as indicating a strong positive effect. After exploration of heterogeneity this finding remains robust although there is significant heterogeneity present on all analyses. There is not enough evidence to determine if this advantage is maintained over time, but the continuation of improvement of the active treatment groups in the trials with longer follow-ups suggest that this was the case.
There was some evidence that TFCBT was a more effective treatment than non-trauma focused therapies (stress management and other therapies). TFCBT was significantly better than other therapies immediately and than stress management at follow-up.
Stress management
There was evidence that stress management was better than wait list/usual care in reducing traumatic stress symptoms and additionally associated symptoms of depression and anxiety although this was based on only 2 studies with a small sample size. There was some evidence that stress management is a more effective treatment than other non-trauma focused therapies, but this was from the results of one study only.
Other therapies
There was no difference between other therapies and waitlist/usual care on the main outcome measure but it did fare better on the self-report traumatic stress and anxiety measures. As stated above other therapies were significantly worse in terms of the primary outcome measure when directly compared with TFCBT and stress management.
Group TFCBT
There was evidence that group TFCBT was better than waitlist/usual care in reducing traumatic stress symptoms although this was based on only one study with a small sample size. There was no difference between group TFCBT and non-trauma focused group CBT.
Trauma focused cognitive behavioural therapy
Eye Movement Desensitisation and Reprocessing
There was evidence that EMDR was better than wait list/usual care in reducing traumatic stress symptoms and additionally associated symptoms of depression and anxiety. The fact that the studies included only small sample sizes and two lacked randomisation concealment means that the results should be interpreted with caution. However, as was the case with TFCBT it is possible that the results may be stronger than suggested by the data, as in several studies the wait list/usual care group received some contact and the expectation that they would be treated which may have been therapeutic. The overall standardised mean difference for clinician rated traumatic stress symptoms post treatment represents a strong positive effect size although the effect size for self-reported PTSD symptom severity did not reach statistical significance.
EMDR appeared to have similar effectiveness to TFCBT in the studies that compared them directly. There was some evidence that EMDR was a more effective treatment than stress management therapies and other therapies.
Anxiety and Depression
Symptoms of anxiety and depression generally improved in line with improvements in traumatic stress symptoms. For treatments such as cognitive restructuring many of the approaches used for PTSD would also be used for anxiety and depression and might explain the improvement. Other treatments such as exposure therapy may also address depressive symptoms through considering issues such as guilt and reponsibility during the processing and through in vivo homework tasks between sessions. However, the treatments may reduce anxiety and depressive symptoms because they are secondary to the PTSD, and when the PTSD improves these symptoms improve. This would suggest that the anxiety and depressive symptoms found in many PTSD sufferers in these studies were secondary to the PTSD rather than being discrete conditions requiring specific treatment.
Adverse effects
Unfortunately no studies reported adverse effects. It is well recognised that adverse effects may occur such as increased reexperiencing following exposure treatment (e.g. Pitman 1991) and the absence of any reporting of them is of major concern.
Dropouts
Most studies reported on dropouts by group which is likely to be contributed to by adverse effects along with other factors. TFCBT and other therapies both did worse than wait list/usual care on this outcome measure but there were no significant differences in drop-out rates in direct comparisons between active treatments. This may reflect the greater logistic demands of treatment versus wait list but may also be as a result of the active treatments not having always been acceptable to those receiving them. This is an important finding and one that should stimulate research to determine the true explanation. If some interventions are not acceptable to those who receive them, the development of interventions that are should be a priority.
Heterogeneity
The Forest plots of the pooled results demonstrated significant heterogeneity between the studies. For example, heterogeneity levels of p < 0.00001 were observed in several analyses of the primary outcome measure. There are likely to be several factors that contribute to the heterogeneity.
There is clearly considerable clinical diversity within the studies considered. An attempt was made to explore this by performing subgroup analyses on the primary outcome measure of TFCBT versus waitlist/usual care. Those studies including only females, all of whom had been sexually or non-sexually assaulted, produced more positive results than the overall results. Possible explanations include the treatments having been superior, females being more responsive to TFCBT than males, traumatisation by assault being more responsive to TFCBT, a combination of these and/or other factors. Those studies that did not include only Vietnam veterans produced a slightly more positive result than all studies. However there was only one study excluded in this subgroup analysis. Therefore the analysis may lack power to show a real difference and great caution must be exercised in interpreting this.
The separation of different active interventions into groups partially addresses the clinical diversity, although not all trials within the same group used identical interventions. The differences were most marked in the "other treatments" group which had in common the absence of cognitive-behavioural techniques and trauma-focused work. There was also diversity in the TFCBT group which included both exposure only and trauma-focused cognitive therapy interventions.
Another source of heterogeneity was the quality of the studies. Sensitivity analyses of higher quality and lower quality studies were performed for the primary outcome measure comparison of TFCBT versus waitlist/usual care to explore this further. The higher quality studies showed better outcomes than the lower quality studies. This finding contradicts previous research (e.g. Moher 1998) that has found an association between poorer methodology and more favourable results for the intervention. Our finding may reflect the fact that the better studies tended to be more recent and associated with refinement of TFCBT techniques. They also included most of the female only studies.
As with all psychological treatment trials there are issues with the control groups. The development of a "psychological treatment placebo" is very difficult, if not impossible, as is blinding of participants and therapists. This can lead to a bias in favour of the active intervention. If present in these studies it would have resulted in the active treatments being likely to appear better than they actually are.
Summary
Thirty-three studies were included in the review. TFCBT and EMDR did significantly better than waitlist/usual care in reducing clinician assessed PTSD symptoms. There was no significant difference between TFCBT, EMDR and SM, although TFCBT and EMDR did significantly better than other therapies. Stress management did significantly better than waitlist/usual care and than other therapies. There was no significant difference between other therapies and waitlist/usual care control. Group TFCBT was significantly better than waitlist/usual care. The considerable unexplained heterogeneity observed in these comparisons, and the potential impact of publication bias on these data, suggest the need for caution in interpreting the results of this review.
Authors' conclusions
Implications for practice 1. Psychological treatment can reduce traumatic stress symptoms in individuals with PTSD. |
Implications for research 1. Further well-designed trials of psychological treatments are required that consider boundary issues (e.g. predictors for treatment effects). The considerable unexplained heterogeneity observed in these comparisons, and the potential impact of publication bias on these data, suggest the need for caution in interpreting the results of this review. |
Acknowledgements
We should like to thank the CCDAN editorial base for their help with searches, helpful comments on the protocol and assistance with the methodology. We should also like to thank the NICE PTSD guideline development group for allowing us access to their datasets.
Data and analyses
- Top of page
- Background
- Objectives
- Methods
- Results
- Discussion
- Authors' conclusions
- Acknowledgements
- Data and analyses
- What's new
- History
- Contributions of authors
- Declarations of interest
- Notes
- Index terms
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||
What's new
Last assessed as up-to-date: 22 May 2007.
| |||||||||
History
Protocol first published: Issue 4, 2001
Review first published: Issue 2, 2005
| |||||||||
Contributions of authors
JIB has been involved in the identification, quality appraisal, data entry, analysis and writing of the review.
MA has been involved in the identification, quality appraisal and and writing of the review.
JIB has been involved in two randomised trials of early psychological interventions designed to prevent PTSD following traumatic events.
Declarations of interest
Nil.
Notes
Eye movement desensitisation and reprocessing (EMDR) was previously removed from the protocol for this review as it was the focus of a separate Cochrane review, EMDR for PTSD. However, the protocol for EMDR for PTSD has now been withdrawn. Trials of EMDR have now been included in this review.
Index terms
Medical Subject Headings (MeSH)
Behavior Therapy [methods]; Cognitive Therapy [methods]; Psychotherapy [*methods]; Psychotherapy, Group; Randomized Controlled Trials as Topic; Stress Disorders, Post-Traumatic [psychology; *therapy]; Stress, Physiological [therapy]
MeSH check words
Adult; Humans
* Indicates the major publication for the study